Container module with an encoder, a recurrent module, and a decoder.

Position-wise Feedforward Layer class FFN(x) = max(0, xW1 + b1) W2+ b2

Trainer class

Positional Encoding class

Position-wise Feedforward Layer class FFN(x) = max(0, xW1 + b1) W2+ b2

Scaled Dot-Product Attention

Optim wrapper for annealing opt

Chinese-English BPE tokenizer. Peculiarities: - Byte-level Byte-Pair-Encoding - Requires a space to start the input string =>

Decoder Layer Transformer class

Decoder Transformer class

Discriminator for adversarial training and multi-task learning

Encoder Transformer class

Encoder Layer Transformer class

Optim wrapper that implements rate.

Container module with an encoder, a recurrent module, and a decoder.

Trainer class

Transformer class args: encoder: Encoder object decoder: Decoder object

Trainer class

Trainer class

Dot product attention. Given a set of vector values, and a vector query, attention is a technique to compute a weighted sum of the val

Trainer class

Trainer class

Position-wise Feedforward Layer Implementation with Convolution class